Well-studied innate immune systems exist throughout bacteria and archaea, but a more recently discovered genomic locus may offer prokaryotes surprising immunological adaptability. Mediated by a cassette-like genomic locus termed Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR), the microbial adaptive immune system differs from its eukaryotic immune analogues by incorporating new immunities unidirectionally. CRISPR thus stores genomically recoverable timelines of virus-host coevolution in natural organisms refractory to laboratory cultivation. Here we combined a population genetic mathematical model of CRISPR-virus coevolution with six years of metagenomic sequencing to link the recoverable genomic dynamics of CRISPR loci to the unknown population dynamics of virus and host in natural communities. Metagenomic reconstructions in an acid-mine drainage system document CRISPR loci conserving ancestral immune elements to the base-pair across thousands of microbial generations. This ‘trailer-end conservation’ occurs despite rapid viral mutation and despite rapid prokaryotic genomic deletion. The trailer-ends of many reconstructed CRISPR loci are also largely identical across a population. ‘Trailer-end clonality’ occurs despite predictions of host immunological diversity due to negative frequency dependent selection (kill the winner dynamics). Statistical clustering and model simulations explain this lack of diversity by capturing rapid selective sweeps by highly immune CRISPR lineages. Potentially explaining ‘trailer-end conservation,’ we record the first example of a viral bloom overwhelming a CRISPR system. The polyclonal viruses bloom even though they share sequences previously targeted by host CRISPR loci. Simulations show how increasing random genomic deletions in CRISPR loci purges immunological controls on long-lived viral sequences, allowing polyclonal viruses to bloom and depressing host fitness. Our results thus link documented patterns of genomic conservation in CRISPR loci to an evolutionary advantage against persistent viruses. By maintaining old immunities, selection may be tuning CRISPR-mediated immunity against viruses reemerging from lysogeny or migration.
Most microbes appear unculturable in the laboratory, limiting our knowledge of how virus and prokaryotic host evolve in natural systems. However, a genomic locus found in many prokaryotes, CRISPR, may offer cultivation-independent probes of virus-microbe coevolution. Utilizing nearby genes, CRISPR can serially incorporate short viral and plasmid sequences. These sequences bind and cleave cognate regions in subsequent viral and plasmid insertions, conferring adaptive anti-viral and anti-plasmid immunity. By incorporating sequences undirectionally, CRISPR also provides timelines of virus-prokaryote coevolution. Yet, CRISPR only incorporates 30–80 base-pair viral sequences, leaving incomplete coevolutionary recordings. To reconstruct the missing coevolutionary dynamics shaping natural CRISPRs, we combined metagenomic reconstructions with population-scale mathematical modeling. Capturing rare and rapid sweeps of CRISPR diversity by highly immune lines, mathematical modeling explains why naturally reconstructed CRISPR loci are often largely identical across a population. Both model and experiment further document surprising proliferations of old viral sequences against which hosts had preexisting CRISPR immunity. Due to these deadly blooms of ancestral viral elements, CRISPR's conservation of old immune sequences appears to confer a selective advantage. This may explain the striking immunological memory documented in CRISPR loci, which occurs despite rapid viral mutation and despite rapid deletions in prokaryotic genomes.